Training Sequence Generation Neural Networks Using Quality Scores

3. The method of claim 1, wherein generating the target likelihood distribution comprises applying a softmax to the highest quality scores for the possible system outputs to generate a respective likelihood for each of the possible system outputs.

4. The method of claim 3, wherein the softmax is applied with reduced temperature.

5. The method of claim 1, wherein the possible system outputs in the vocabulary comprise tokens in a natural language.

6. The method of claim 1, wherein the quality score is based on the edit distance between the candidate output sequence and the ground truth output sequence, and wherein the highest quality score that can be assigned is the quality score for the candidate output sequence that has a smallest edit distance to the ground truth output sequence.

10. The system of claim 9, wherein generating the target likelihood distribution comprises applying a softmax to the highest quality scores for the possible system outputs to generate a respective likelihood for each of the possible system outputs.

11. The system of claim 10, wherein the softmax is applied with reduced temperature.

12. The system of claim 9, wherein the possible system outputs in the vocabulary comprise tokens in a natural language.

13. The system of claim 9, wherein the quality score is based on the edit distance between the candidate output sequence and the ground truth output sequence, and wherein the highest quality score that can be assigned is the quality score for the candidate output sequence that has a smallest edit distance to the ground truth output sequence.

17. The non-transitory computer-readable storage media of claim 16, wherein generating the target likelihood distribution comprises applying a softmax to the highest quality scores for the possible system outputs to generate a respective likelihood for each of the possible system outputs.

18. The non-transitory computer-readable storage media of claim 16, wherein the quality score is based on the edit distance between the candidate output sequence and the ground truth output sequence, and wherein the highest quality score that can be assigned is the quality score for the candidate output sequence that has a smallest edit distance to the ground truth output sequence.